European Journal of Endocrinology - accepted 26 May 2016

Does metabolic health in overweight and obesity persist? - Individual variation and

cardiovascular mortality over two decades

Akaal Kaur, Desmond G. Johnston, Ian F Godsland.

Diabetes Endocrinology and Metabolic Medicine, Faculty of Medicine, Imperial College

London, St. Mary's Campus, London, UK

Short title: metabolically healthy overweight and obesity

Key terms: metabolic syndrome, insulin resistance, individual variation, cardiovascular

disease

Word count: 4709

Corresponding author and person to whom reprint requests should be addressed:

Ian F Godsland, PhD,

Division of Diabetes, Endocrinology and Metabolism, Department of Medicine,

Imperial College London, St Mary’s Campus,

Room G1, Norfolk Place, London W2 1NH, UK.

email: i.godsland@imperial.ac.uk

Abstract

Objective: Overweight and obese individuals may be metabolically healthy but attention

needs to be given to long-term persistence of this trait and any associated variation in

cardiovascular risk.

Design: Cross-sectional and longitudinal variation in metabolic health and associated

cardiovascular mortality were analysed in 1099 white European-origin normal weight and

overweight or obese males followed for 20 years.

Methods: Definitions of metabolic health were based on LDL and HDL cholesterol,

triglycerides, blood pressure, fasting glucose and cardiovascular risk. Insulin resistance (e.g.

HOMA-IR) and subclinical inflammation (ESR and white blood cell count) were explored.

Cardiovascular mortality risks and persistence of metabolic health status were evaluated.

Results: There were 87 cardiovascular deaths. Insulin resistance was increased in

metabolically healthy overweight or obese participants (median HOMA-IR 2.63, 95%ci 1.79-

3.65, p<0.001) relative to normal weight (median HOMA-IR 1.67, 95%ci 1.08-2.67, p<0.001)

as was subclinical inflammation but metabolically healthy overweight or obese individuals

were not at increased risk of cardiovascular mortality compared with the metabolically

healthy normal-weight (hazard ratio 1.13, 95% ci 0.34-3.72, p=0.8). The proportions of

initially metabolically healthy overweight or obese who remained metabolically healthy for

visits 2, 3 and 4 were 54, 48 and 39%, respectively, and for initially normal-weight

individuals, 68, 51 and 41%. A lower proportion of metabolically healthy overweight or obese

individuals remained metabolically healthy at visit 2 compared with normal weight individuals

(p=0.007) but proportions converged thereafter..

Conclusions: Despite being insulin resistant and having greater subclinical inflammation, and

despite instability in metabolic health status, metabolically healthy overweight or obese

individuals were at no greater risk of cardiovascular mortality than their normal-weight

equivalents.

Introduction

Overweight (OW) and obesity (OB) are associated with physiologic and metabolic

disturbances that predispose towards the development of type 2 diabetes mellitus,

cardiovascular disease (CVD) and cancer. In the 1980s, subgroups of OB individuals were

identified that appeared free of typical adiposity-related disturbances1 and it was suggested

that some OB individuals might not be at increased risk2. Subsequently, the concept of

metabolically healthy (MH) obesity was adopted3. Estimates of the prevalence of MH OB in

Europe range between 2-19 percent in men and 7-28 percent in women4.

Problems consistently raised in commentaries on MH OW and OB have included the

lack of an agreed definition of metabolic health and the fact that a continuum of variation in

metabolic disturbance extends across all levels of adiposity5-8. Nevertheless, if a pragmatic

definition of MH could be agreed, it might yet be used to identify OW or OB individuals who

need not benefit from intensive weight loss. In a detailed review, Stefan and colleagues

distinguished three types of investigation, differing according to whether MHO had been

defined with reference to insulin sensitivity, cardiorespiratory fitness or a putative ‘metabolic

syndrome’7. Measures of insulin sensitivity and cardiorespiratory fitness (and inflammatory

markers, which have also been used to define MH) are not routinely available. However,

defining MH with reference to readily-available risk factor measures does offer a practical

tool with which low risk OW or OB individuals might be identified.

In the present analysis, we explore data from the HDDRISC study, in which a broad

range of risk factors was recorded in 1191 white European males, a substantial proportion of

whom had repeated follow-up information recorded at 2-3 year intervals. Applying a

definition of MH based on readily-available risk factor measurements and their associated

CVD mortality risks in the HDDRISC cohort, we test the following hypotheses: 1) MH

overweight or obese (OWOB) individuals are at greater risk of CVD mortality than MH

normal weight (NW) individuals; 2) MH OWOB individuals are insulin resistant and exhibit

subclinical inflammation relative to MH NW individuals and 3) MH in OWOB individuals is a

persistent trait, sustained over long-term follow-up.

Subjects and Methods

Design

The Heart Disease and Diabetes Risk Indicators in a Screened Cohort (HDDRISC)

Study began as a company health program to evaluate risk factors for CVD and diabetes.

The study recruited 1191 white European males (age range 25-70y; BMI range 18.1-46.9

kg/m2) who were generally healthy and either in employment or recently retired and data

acquisition was between June 1971 and January 2000. Metabolic, clinical and laboratory

measurements were recorded for each participant at their first visit and, for those continuing

in the study, at subsequent follow-up visits every 2-3 years. Ethics committee approval was

obtained and each participant gave written, informed consent.

Procedures

Procedures have been described in detail previously9. Briefly, participants were

instructed to fast overnight and, prior to attendance at a metabolic day ward the next

morning, refrain from drinking alcohol and smoking. Demographic data, including self-

reported cigarette smoking (0, <15, ≥15 cigarettes per day), alcohol consumption (0, <28, ≥

28 units per week) and exercise habits (none, nonaerobic, aerobic, with aerobic

distinguished as 3 or more periods of ≥20 minutes of exercise to breathlessness) and

medical history were recorded. After 15 minutes in a semi-recumbent position, systolic and

diastolic blood pressure (SBP and DBP, respectively) were measured. Blood samples were

taken from an indwelling cannula for laboratory measurements. A further sample was taken

5-10 minutes later for a second measurement of fasting plasma glucose and insulin (FPG

and FPI). The majority of participants then underwent an oral glucose tolerance test (OGTT:

50% dextrose solution at 1g glucose/kg body weight) and samples were taken to measure

plasma glucose and insulin at 30, 60, 90, 120, 150 and 180 minutes.

Laboratory Measurements

Plasma glucose and insulin and serum triglycerides and total, HDL and LDL

cholesterol concentrations were measured by the HDDRISC research group as previously

described9. Routine haematology and biochemistry measurements included white blood cell

count (WBC) and erythrocyte sedimentation rate (ESR) as indices of subclinical

inflammation and were undertaken either by the research group or the local chemical

pathology laboratory. Between- and within- batch quality control was monitored by frozen,

pooled plasma samples and commercially obtained lyophilised sera and by participation in

relevant national quality assurance schemes with particular attention paid to long-term assay

standardisation.

Definition of Metabolic Health

Clearly, increased BMI cannot be included as a criterion for MH in an analysis of MH

OW or OB and this extends to increased waist circumference which displays marked co-

linearity with BMI4 and has been excluded from the majority of definitions of MH OW or OB7,

10-12. Other readily-available criteria then include plasma lipid, lipoprotein and glucose

concentrations and blood pressure. Five criteria are considered in the present analysis: LDL-

C of ≥3.36 mmol/l or use or a statin and (with cut-offs defined according to metabolic

syndrome criteria13, 14) TG ≥1.7 mmol/l or use of a fibrate, HDL-C <1.03 mmol/l, SBP ≥130

mmHg systolic or DBP ≥85 mmHg or use of blood pressure-lowering agents and FPG ≥5.6

mmol/l or a diagnosis of diabetes mellitus and/or use of glucose-lowering agents. The

question then arises, how many, if any, of the criteria may be allowed to be present before

MH can be excluded? Previous studies of MHO employing similar criteria have allowed none

4, 12, 15, ≤1 10, 11, 16, 17 or ≤2 18-21. In the present analysis, in the full dataset prior to exclusions

(see below), each of the 5 risk factors was confirmed as a predictor of CVD and total

mortality in univariate Cox proportional hazards models, unadjusted for any factors that

might influence variation in risk factor levels. Participants were then distinguished according

to whether they had 0, 1, 2, 3, 4 or 5 risk factors and numbers of risk factors were entered as

dummy variables in an unadjusted Cox model to determine at what level of risk factor

number significant risk of CVD or total mortality would become apparent. Participants with

fewer than that number were then identified as MH - ‘metabolically healthy’ - and those with

that number or more as MUH - ‘metabolically unhealthy’. We, therefore, ‘calibrated’ our

criteria for MH against actual clinical risks they conferred in the HDDRISC cohort.

Definition of overweight

Preliminary analysis indicated that among obese (BMI ≥30.0 kg/m2) participants,

numbers were insufficient for a reliable evaluation of CVD mortality (12 CVD deaths among

the 83 OB participants, with only 1 CVD death (5 deaths in total) among the 13 MH OB

participants). Therefore, OW and OB individuals were combined in a single group (OWOB).

Rather than the conventional lower limit for OW of 25.0 kg/m2, we adopted a limit of 27.0

kg/m2 to enable a meaningful evaluation of the relative benefits or otherwise of MH in

overweight individuals: with a lower limit of BMI of ≥25.0 kg/m2, the Cox model hazard ratio

for OWOB as a predictor of CVD mortality (101 CVD deaths prior to exclusions) was 1.32,

95%ci 0.90-1.94, p=0.1 whereas with a lower limit of BMI of ≥27.0 kg/m2, the hazard ratio

was 1.95, 95%ci 1.33-2.87, p=0.001. A lower limit of 27 kg/m2 therefore provided for an

appreciably higher risk background against which to distinguish whether MH OWOB

individuals were or were not at increased risk relative to MH NW individuals.

Cardiovascular disease and total mortality ascertainment

Deaths in the U.K. were identified through the NHS Information Centre for Health and

Social Care. Deaths overseas were notified by family members of the deceased and by the

company for which the health screening programme was undertaken. Mortality information

was complete to 1st January 2014, with the exception of 27 individuals who are no longer

traceable. Fifteen of the 27 who were lost to mortality follow-up were also lost to any further

clinical follow-up after their first study visit and therefore had zero follow-up time. Mean

mortality follow up time in the remaining 1176 was 25.7 years (range 0.5-42.5 years).

Death from CVD was assigned if myocardial infarction, coronary artery disease or

thrombosis, cerebrovascular incident, stroke, peripheral vascular disease or aortic aneurysm

(ICD-10 codes I20-I25; I63-I67 and I70-I73) was recorded as a primary cause of death.

Data Analysis

Participants with pre-existing CVD or incomplete data for BMI, blood pressure,

triglycerides, HDL-C, FPG and LDL-C were excluded. FPG and FPI concentrations were

calculated as the mean of the two fasting measurements. The homeostasis model

assessment index of insulin resistance (HOMA-IR)22 and the Matsuda index of insulin

sensitivity (Matsuda-SI)23 were calculated.

Statistical analyses were carried out using StataCorp. 2013 (Stata Statistical

Software: Release 13. College Station, TX: StataCorp LP). Risks of CVD and total mortality

in MH or MUH OWOB participants relative to their NW equivalents were evaluated by Cox

proportional hazards modelling with adjustment for age, smoking, alcohol and exercise (as

categorised above) to diminish any confounding effect these covariates might have on

relationships between adiposity, metabolic health and CVD mortality risk. Primary analyses

concerned relationships between baseline characteristics and CVD and mortality risks.

Secondary analyses included follow-up information, with OWOB and confounding variables

treated as discrete, time-varying covariates. Kruskal-Wallis analysis of variance, the Mann-

Whitney test and the proportions test were applied as appropriate. The number of

participants decreased with increasing number of follow-up visits. Therefore, analyses were

undertaken for visits 1 to 2 in the subgroup of individuals with 2 or more visits, for visits 1 to

3 in the subgroup of individuals with 3 or more visits, through to visits 1 to 10 in the subgroup

of individuals with 10 or more visits. In this way, numbers of observations at each visit within

each number-of-visits subgroup were the same. Risk of transition from MH to MUH was

evaluated by Cox proportional hazards modelling in three models according to the ‘naïve’

and ‘longitudinal’ approaches distinguished by Twisk et al24. In Model 1 time to first

occurrence of MUH in those MH at baseline was predicted by baseline OWOB status and

covariates. In Models 2 and 3, MUH was treated as a recurrent event in the full dataset and

predicted, according to the Andersen-Gill model25, by OWOB status and covariates either at

baseline (Model 2) or as discrete, time-varying covariates (Model 3).

Results

Characteristics at first visit for the 1191 participants in the cohort are summarised in

Table 1. Exclusions due to existing CVD or missing metabolic health variables are

summarised in Supplementary Figure 1. After exclusions, there were 1099 participants with

a first visit with complete data for BMI, LDL-C, TG, HDL-C, blood pressure and FPG. Median

age was 46.6y, (range 25-70y) and median BMI 25.6kg/m2 (range 18.1-46.9kg/m2). Sixty-

nine percent were non-smokers; 65% consumed <28 units/week of alcohol; 40% took no

exercise, 41% undertook nonaerobic exercise. Prescription drug use at baseline was limited:

5 on glucose-lowering agents, 3 on statins, 1 on a fibrate and 20 on blood pressure-lowering

agents. Of the 1099 participants, 955 had a measurement of FPI as well as FPG and 631

had an OGTT with measures of plasma glucose and insulin.

Criteria for metabolic health

After exclusions, there were 87 CVD deaths during a mean follow-up time of 24.6

years (range 0.5-42.5 years, excluding 9 participants lost to mortality follow-up and with no

clinical follow-up). In univariate Cox proportional hazards modelling, each of the five defining

risk factors for MH was a significant predictor of CVD and of total mortality (Supplementary

Table 1). Nevertheless, in the presence of only a single risk factor of any type, risks of CVD

and of total mortality (unadjusted for factors contributing to risk factor variation) were not

significantly raised (hazard ratio (HR) CVD mortality 1.66, 95% ci 0.62-4.48, p=0.3; total

mortality 1.51, 95% ci 0.95-2.40, p=0.08). With two risk factors, however, the hazard ratio

became significant (HR CVD mortality 2.79, 95%ci 1.07-7.27, p=0.03; total mortality 2.07,

95%ci 1.31-3.28, p=0.002). Therefore, our definition of MH allowed for ≤1 of the 5 risk

factors considered. Fifty-four percent of NW and 28% of OWOB individuals were identified

as MH (Table 2). MH OWOB participants tended to smoke and drink more than MH NW.

Despite being classified as MH according to the specified cut-offs, MH OWOB participants

tended to have higher FPG, blood pressure, triglycerides and uric acid and lower HDL

cholesterol than MH NW and similar differentials were apparent between MUH OBOW and

MUH NW participants. Among OBOW participants, those who were MH were slightly

younger and took more exercise than the MUH, but tended to smoke more and have higher

alcohol intake.

Insulin sensitivity and subclinical inflammation in metabolically healthy NW and OWOB

MH OWOB participants had markedly higher insulin resistance than MH NW, 57%

higher according to HOMA-IR and 56% higher according to Matsuda-IS (Table 2). Similar

differentials were also seen between MUH OWOB and NW participants. White cell count

was also significantly increased in MH OWOB participants compared with MH NW but this

difference was not apparent in MUH participants. ESR showed relatively minor variation,

however.

Metabolic health and CVD and total mortality in NW and OWOB

Independently of age, smoking, alcohol intake and exercise habit, OWOB was a

significant predictor of both CVD (HR 1.76, 95%ci 1.26-2.44, p=0.001) and total (HR 1.38,

95%ci 1.13-1.67, p=0.001) mortality. However, among those who were MH, OWOB ceased

to be an independent predictor of CVD mortality (HR 1.13, 95% ci 0.34-3.72, p=0.8). An

effect on total mortality was, nevertheless, still apparent (HR 1.60, 95% ci 1.04-2.47,

p=0.03). In the full dataset (3129 observations), with OWOB and confounding factors treated

as discrete, time-varying covariates, OWOB was again a significant, independent predictor

of both CVD (HR 2.02, 95%ci 1.27-3.21, p=0.003) and total (HR 1.55, 95%ci 1.20-2.01,

p=0.001) mortality. Among those who were MH at baseline, however, OWOB was neither a

risk factor for CVD (HR 0.64, 95%ci 0.15-2.83, p=0.5) nor for total (HR 1.38, 95% ci 0.83-

2.90, p=0.2) mortality. In contrast to those who were MH, among participants who were MUH

at baseline, OWOB remained an independent predictor of mortality with hazard ratios of 1.67

(95%ci 1.17-2.39, p=0.004) and 1.26 (95% ci 1.01-1.57, p=0.03) for CVD and total mortality,

respectively, and these risks were equally apparent when the full dataset with covariate

follow-up was considered.

Prevalence of MH during follow-up

Of the 1099 participants included in this analysis, 608 had a second visit, 404 a third

down to 32 with 10 visits. Mean follow-up time between visits 1 and 2 was 3.7 years and

ranged between subsequent visits from 2.1 to 2.7 years. Of the 316 participants who were

OWOB at their first visit, 160 had one subsequent visit, 96 two subsequent visits to 9 with 10

subsequent visits.

Baseline prevalence of MH was 54% among the 783 NW participants and 28%

among the 316 OWOB participants (Supplementary Table 2). In the subgroup of 32

participants who returned for 10 or more visits the baseline prevalence of MH was 39% in

NW and 44% in OBOW individuals and at visit 10, 33% and 38%. In OWOB individuals,

there was a trend towards increased prevalence of MH from visits 1 to 3 (average

prevalences across number-of-visits subgroups, weighted by numbers of observations in

each subgroup, were 30, 35, 43% for visits 1-3, respectively), with an irregular decline in

prevalence thereafter. Among individuals OWOB at baseline, baseline prevalences of MH in

individuals in number-of-visits subgroups 1-10 were 28, 26, 31, 35, 33, 35, 38, 34, 41 and

44%, respectively, indicating that those who subsequently returned for increasing numbers

of follow-up visits had an increased prevalence of MH from the outset.

Changes in MH status during follow-up

A constant prevalence from one visit to the next might be sustained by equal

numbers transitioning from MH to MUH and from MUH to MH. Therefore, we analysed

proportions of individuals MH at baseline who remained MH at successive visits. Proportions

of individuals MH and OWOB at baseline who remained MH at visit 2 varied between 49 and

60% for number-of-visits subgroups 1-8 (subgroups 9 and 10 having unreliably small

numbers of observations: n<10), with little evidence of any trend (Table 3). For NW

individuals, the range for number-of-visits subgroups 1-10 was between 60% and 71%.

Weighted average proportions of individuals OWOB and MH at baseline who remained MH

at visits 2, 3 and 4 were 54, 48 and 39%, respectively. From visit 5 onwards proportions

remaining MH continued to decline, but, overall, only the difference in proportions between

visits 1 and 2 was statistically significant. Proportions of individuals NW and MH at baseline

who remained MH at visits 2, 3 and 4 were 68, 51 and 41%, with the difference in

proportions between visits 1 and 2 being generally significant at p<0.001 and between visits

2 and 3, at p<0.05. Initially, the rate of attrition of individuals remaining MH was faster

among individuals OWOB at baseline than among those NW (Figure 1) and, for those with

≥2 visits, the proportion of OWOB individuals remaining MH at visit 2 was significantly lower

than the proportion of NW individuals remaining MH (p=0.007). However, beyond visit 2

proportions of OWOB and NW participants remaining MH converged. For those with ≥3

visits, analysis of individuals remaining MH was repeated with visit 2 as baseline and, for

those with ≥4 visits, visit 3 as baseline. Findings were generally consistent with the analysis

with an individual’s visit 1 as baseline (Table 4).

Changes in weight and lifestyle during follow-up

For those with ≥2 visits, 29% of those OWOB at visit 1 were NW at visit 2. For those

with ≥3 visits, 15% of those OWOB at visit 2 were NW at visit 3. For those with ≥4 visits,

23% of those OWOB at visit 3 were NW at visit 4. Equivalent figures for NW individuals

becoming OWOB were 6, 8 and 7%. For visits 5 to 10, proportions ranged between 6-16%

for OWOB switching to NW and 0-9% for NW to OWOB. Transition from OWOB to NW

tended to be associated with transition from MUH to MH, although the majority did not

change status (results not shown) . For the transition from NW to OWOB, again, the majority

did not change status. Transitions in smoking, alcohol consumption and exercise were also

explored (results not shown). There was a tendency towards reductions in smoking over

successive visits but with no clear association with change in MH status. On Cox

proportional hazards modelling of incident MUH in the dataset including follow-up visits, only

increasing age was a consistent predictor of transition to MUH in those MH at baseline

(Model 1: HR1.06, 95%ci 1.03-1.08, p<0.001; Model 2: HR 1.05, 95%ci 1.03-1.07, p<0.001;

Model 3: HR 1.03, 95%ci 1.01-1.04, p<0.001). There was some evidence for moderate

exercise reducing risk of transition to MUH (Model 1: HR 0.70, 95%ci 0.47-1.02, p=0.06;

Model 2: HR 0.73, 95%ci 0.57-0.93, p=0.01) and for heavy smoking increasing risk (Model 3:

HR 1.47, 95%ci 0.99-2.16, p=0.05) but, importantly, there was no evidence from any of

these analyses that among those MH at baseline those who were OWOB were at an overall

increased risk of becoming MUH subsequently.

Discussion

We evaluated three hypotheses. The first was that MH OWOB individuals are at

greater risk of CVD mortality than MH NW individuals. This was not the case. Other

studies11, 26 have noted that MH OWOB individuals tend to appear somewhat healthier in

terms of weight and lifestyle characteristics than their MUH counterparts, although in our

study this was only apparent with regard to exercise; smoking and alcohol intake tended to

be greater in the MH OWOB participants. Overall, however, risk differentials were

independent of lifestyle covariates, which accords with the possibility that among OWOB

individuals a low CVD risk subgroup can indeed be distinguished using MH criteria and

supports the concept of metabolically healthy overweight or obesity and this low risk may

extend to total mortality.

The second hypothesis was that MH OWOB individuals are insulin resistant and

exhibit subclinical inflammation relative to MH NW individuals. This was the case and, with

regard to inflammation, accords with a previous report of increased inflammation in MHO

individuals27. This could suggest that further discrimination of MH among OWOB individuals

may be possible using measures of insulin resistance and subclinical inflammation.

However, our finding that MH OWOB individuals were at no greater risk of CVD mortality

than MH NW individuals suggests that such further discrimination may add little to CVD risk

evaluation, consistent with CVD risk associated with increasing adiposity being fully

explained by established risk factors28. However, although increased subclinical

inflammation in MHO may be relatively benign with regard to CVD risk, the increased insulin

resistance may indicate increased risk of type 2 diabetes26, 29, 30.

Our third hypothesis was that MH in OWOB individuals is a persistent trait, sustained

over long-term follow-up. Against this, from visit 1 to 2, fully 51% of MH individuals who were

OBOW and 29% of those who were NW and at visit 1 were MUH at visit 2. Subsequently,

over two decades of follow-up and with up to 10 visits, only about 30% of individuals either

OBOW or NW and MH at baseline remained consistently MH. Moreover, similar percentages

transitioned from MUH to MH over the course of these visits (results not shown). That MH is

susceptible to this degree of variability does not appear to have been realised in previous

studies of MHO or MH OWOB.

Our long-term MH profiles therefore indicate appreciable intra-individual variation in

MH status, and this would be expected from the individual variation exhibited by the

variables used to categorise MH status31. Use of a categorical criterion, deriving from a

continuous measure of a biologically variable parameter, measured on a single occasion to

classify individuals as having or not having a condition is highly susceptible to regression

towards the mean and, judging by the marked attrition in MH individuals from visit 1 to 2, this

had considerable impact on MH categorisation. Beyond visit 2, however, the rate of attrition

of individuals consistently MH was reduced somewhat. Importantly, however,, in survival

analysis with incident MUH as outcome, among those who were MH at baseline, OWOB was

entirely unrelated to transition MUH; moreover, transition to MUH could not be accounted for

by weight gain and was only associated with lifestyle change to a limited extent, consistent

with MH at baseline being, to some extent, a stable characteristic. We may, therefore, infer

that each participant in the study had an underlying degree of metabolic health that was

further from or closer to the categorical cut-offs applied to define MH. Those far from the cut-

offs would have tended to retain MH or MUH status, whereas those close to the cut off would

have transitioned from one state to another over time. It is noteworthy, however, that the

proportion of individuals transitioning from MH to MUH from visit 1 to 2 was significantly

greater among OWOB than among NW individuals. Possibly, given that MH was less

prevalent among OWOB individuals, classifications of MH at baseline might be more

susceptible to greater extremes of variation, which would be more likely to relapse into MUH

at a second visit.

A number of previous studies have evaluated CVD risk in MH OB and, in accord with

our findings, and despite inconsistencies in the definitions of MH, the majority have found

CVD risk not to be increased in MH OB compared to MH NW individuals18, 26, 29, 32-35, or have

found CVD risk be to lower in MH OB compared with MUH OB36, but not all studies agree10.

Moreover, several studies have found evidence for increased risk in MH OB compared to

MH NW individuals in studies with >15 years follow-up17, 19, 37. However, our findings, which

concerned CVD mortality over >20 years of follow-up, did not bear this out.

Few previous studies have included a follow-up evaluation of MH status. Appleton

and colleagues undertook baseline and follow-up evaluation in a population sample of 3743

individuals over a median of 8.2 years with 61% providing clinic data at follow-up26. In

agreement with our findings, a significantly higher proportion of MH OB than MH NW

individuals were MUH at a second visit and risk of CVD in the MH OB was no greater than in

MH NW individuals. Also in accord with our findings, the proportion of individuals MHO at

baseline becoming MUH at follow up was substantial at 30%. Likewise, Soriguer and

colleagues, in a study of 1051 individuals found that 48% of those MH OB at baseline were

MUH at a 6 year follow-up30. However, in neither of these studies was the time profile of

attrition of MH distinguished, there being only a single follow-up estimate of change in MH

status.

A limitation of our study was the relatively small sample size, potentially exacerbated

by exclusion of 92 participants due to missing data. Analysis of the dataset with imputation of

missing values resulted in identical findings with regard to CVD mortality (results not shown).

However, with the imputed dataset, having 1 risk factor was associated with increased risk of

total mortality. Further sub-analysis with MH defined by having no risk factors confirmed

relative protection for total mortality among MH OWOB individuals, but there were no CVD

deaths among OWOB participants with no risk factors. It should be acknowledged that the

non-significant hazard ratio of 1.66 we observed for CVD mortality among those with a single

risk factor could well become statistically significant with greater numbers, so our definition

of metabolic health must be regarded as relative and no more than a working definition. It

should also be acknowledged that allowing for a single risk factor in our definition of MH

could lead to inclusion of people with diabetes as metabolically healthy. This might seem

unjustified; however, it should be borne in mind that a participant with diabetes as a sole risk

factor would have to have been free of abnormalities in all other risk factors and not be

taking lipid- or blood pressure-lowering agents. In any case, exclusion of people with

diabetes from our analysis made no difference to our findings (results not shown). The small

numbers of clinical outcomes meant that we could not systematically analyse CVD risk

according to change in MH status, as has been done in one previous study, which was able

to demonstrate that risk of CVD in the MHO was no greater than in MH NW individuals

regardless of subsequent transition to MUH26. Nevertheless, in accord with this, we could

conclude, that despite appreciable transitioning between MH and MUH, OBOW individuals

who were MH at baseline were not at increased risk. Limitations also included the highly

selected group, which limits generalizability, although its homogeneity represents a strength

with respect to hypothesis testing. Another limitation was the progressive loss to follow-up

with regard to metabolic testing. It should be noted that our follow-up was undertaken in the

context of a health program and there was evidence for weight loss in the cohort overall.

Weight loss might have been expected to reduce transitioning from MH to MUH and possibly

contributed to an increase in the prevalence of MH among individuals OBOW at baseline up

to visit 3.

In conclusion, our findings indicate appreciable individual variability and regression

towards the mean when MH status is assigned by widely-used criteria. MH among OW and

OB individuals was associated with increased insulin resistance and subclinical inflammation

compared with MH NW individuals. Nevertheless, despite these differences, OWOB

individuals classified as MH at baseline according to either definition of MH were at no

greater risk of CVD mortality over two decades of follow-up than MH NW individuals,

supporting the possibility that a discrete category of metabolically healthy overweight or

obesity can be distinguished.

Declaration of interest

The authors have no interests to declare

Funding

Data acquisition for the HDDRISC study was funded by the Heart Disease and Diabetes

Research Trust and the Rosen Foundation.

Author contributions

AK: statistical analysis, manuscript writing; DGJ: analysis plan and manuscript

development, IFG analysis plan, statistical analysis, manuscript writing

Acknowledgements

The HDDRISC study was initiated the late Professor Victor Wynn, who directed it for much

of its course and was funded by the Heart Disease and Diabetes Research Trust and the

Rosen Foundation. Data acquisition was sustained by many clinical, scientific, technical,

nursing and administrative staff, to each of whom we would like to extend our thanks.

References

1. Ruderman NB, Schneider SH & Berchtold P. The "metabolically-obese," normal-

weight individual. Am J Clin Nutr 1981 34 1617-1621.

2. Andres R. Effect of obesity on total mortality. Int J Obes 1980 4 381-386.

3. Sims EA. Are there persons who are obese, but metabolically healthy? . Metabolism

2001 50 1499-1504.

4. van Vliet-Ostaptchouk JV, Nuotio ML, Slagter SN, Doiron D, Fischer K, Foco L, Gaye

A, Gögele M, Heier M, Hiekkalinna T, Joensuu A, Newby C, Pang C, Partinen E,

Reischl E, Schwienbacher C, Tammesoo ML, Swertz MA, Burton P, Ferretti V,

Fortier I, Giepmans L, Harris JR, Hillege HL, Holmen J, Jula A, Kootstra-Ros JE,

Kvaløy K, Holmen TL, Männistö S, Metspalu A, Midthjell K, Murtagh MJ, Peters A,

Pramstaller PP, Saaristo T, Salomaa V, Stolk RP, Uusitupa M, van der Harst P, van

der Klauw MM, Waldenberger M, Perola M & Wolffenbutte lBH. The prevalence of

metabolic syndrome and metabolically healthy obesity in Europe: a collaborative

analysis of ten large cohort studies. BMC Endocr Disord. 2014 Feb 1;14:9. doi:

10.1186/1472-6823-14-9 2014 14 9.

5. Pataky Z, Bobbioni-Harsch E & Golay A. Open questions about metabolically normal

obesity. Int J Obes (Lond) 2010 34 Suppl 2 S18-23.

6. Phillips CM. Metabolically healthy obesity: definitions, determinants and clinical

implications. Rev Endocr Metab Disord 2013 14 219-227.

7. Stefan N, Häring HU, Hu FB & Schulze MB. Metabolically healthy obesity:

epidemiology, mechanisms, and clinical implications. Lancet Diabetes Endocrinol

2013 1 152-162.

8. Blüher M. Are metabolically healthy obese individuals really healthy? Eur J

Endocrinol. 2014 171 R209-219.

9. Godsland IF, Leyva F, Worthington M, Walton C & Stevenson JC. Associations of

smoking, alcohol and physical activity with risk factors for coronary heart disease and

diabetes in the first follow-up cohort of the HDDRISC Study (HDDRISC-1). J Intern

Med 1998 244 33-41.

10. Kuk JL & Ardern CI. Are metabolically normal but obese individuals at lower risk for

all-cause mortality? Diab Care 2009 32 2297-2299.

11. Ortega FB, Lee DC, Katzmarzyk PT, Ruiz JR, Sui X, Church TS & Blair SN. The

intriguing metabolically healthy but obese phenotype: cardiovascular prognosis and

role of fitness. Eur Heart J 2013 34 389-397.

12. Marini MA, Frontoni S, Succurro E, Arturi F, Fiorentino TV, Sciacqua A, Perticone F

& Sesti G. Differences in insulin clearance between metabolically healthy and

unhealthy obese subjects. Acta Diabetol 2014 51 257-261.

13. NCEP. Executive summary of the Third Report of the National Cholesterol Education

Program (NCEP) Expert Panel on Detection, Evaluation and Treatment of High Blood

Cholesterol in Adults (Adult Treatment Panel III). JAMA 2001 285 2486-2497.

14. Alberti KG, Zimmet P, Shaw J & IDF-Epidemiology-Task-Force-Consensus-Group.

The metabolic syndrome - a new worldwide definition. Lancet 2005 366 1059-1062.

15. Vukovic R, Milenkovic T, Mitrovic K, Todorovic S, Plavsic L, Vukovic A & Zdravkovic

D. Preserved insulin sensitivity predicts metabolically healthy obese phenotype in

children and adolescents. Eur J Pediatr 2015 Jul 5. [Epub ahead of print].

16. Wildman RP, Muntner P, Reynolds K, McGinn AP, Rajpathak S, Wylie-Rosett J &

Sowers MR. The obese without cardiometabolic risk factor clustering and the normal

weight with cardiometabolic risk factor clustering: prevalence and correlates of 2

phenotypes among the US population (NHANES 1999-2004). Arch Intern Med 2008

168 1617-1624.

17. Hinnouho GM, Czernichow S, Dugravot A, Batty GD, Kivimaki M & Singh-Manoux A.

Metabolically healthy obesity and risk of mortality: does the definition of metabolic

health matter? Diab Care 2013 36 2294-2300.

18. Kip KE, Marroquin OC, Kelley DE, Johnson BD, Kelsey SF, Shaw LJ, Rogers WJ &

Reis SE. Clinical importance of obesity versus the metabolic syndrome in

cardiovascular risk in women: a report from the Women's Ischemia Syndrome

Evaluation (WISE) study. Circulation 2004 17 706-713.

19. Ärnlöv J, Ingelsson E, Sundström J & Lind L. Impact of body mass index and the

metabolic syndrome on the risk of cardiovascular disease and death in middle-aged

men. Circulation 2010 121:230-236.

20. Durward CM, Hartman TJ & Nickols-Richardson SM. All-cause mortality risk of

metabolically healthy obese individuals in NHANES III. J Obes 2012 2012 460321.

21. Choi KM, Cho HJ, Choi HY, Yang SJ, Yoo HJ, Seo JA, Kim SG, Baik SH, Choi DS &

Kim NH. Higher mortality in metabolically obese normal-weight people than in

metabolically healthy obese subjects in elderly Koreans. Clin Endocrinol (Oxf) 2013

79 364-370.

22. Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF & Turner RC.

Homeostasis model assessment: insulin resistance and ß-cell function from fasting

plasma glucose and insulin concentrations in man. Diabetologia 1985 28 412-419.

23. Matsuda M & DeFronzo RA. Insulin sensitivity indices obtained from oral glucose

tolerance testing. Diabetes Care 1999 22 1462-1470.

24. Twisk J, Smidt N & de Vente W. Applied analysis of recurrent events: a practical

overview. J Epidemiol Community Health 2005 59 706.

25. Andersen P & Gill R. Cox’s regression model for counting processes: A large sample

study. Annals of Statistics 1982 10 1100–1120.

26. Appleton SL, Seaborn CJ, Visvanathan R, Hill CL, Gill TK, Taylor AW, Adams RJ,

North, West, Adelaide, Health, Study & Team. Diabetes and cardiovascular disease

outcomes in the metabolically healthy obese phenotype: a cohort study. Diab Care

2013 36 2388-2394.

27. Gómez-Ambrosi J, Catalán V, Rodríguez A, Andrada P, Ramírez B, Ibáñez P, Vila N,

Romero S, Margall MA, Gil MJ, Moncada R, Valentí V, Silva C, Salvador J &

Frühbeck G. Increased cardiometabolic risk factors and inflammation in adipose

tissue in obese subjects classified as metabolically healthy. Diab Care 2014 37 2813-

2821.

28. Emerging, Risk, Factors, Collaboration, Wormser D, Kaptoge S, Di Angelantonio E,

Wood AM, Pennells L, Thompson A, Sarwar N, Kizer JR, Lawlor DA, Nordestgaard

BG, Ridker P, Salomaa V, Stevens J, Woodward M, Sattar N, Collins R, Thompson

SG, Whitlock G & J D. Separate and combined associations of body-mass index and

abdominal adiposity with cardiovascular disease: collaborative analysis of 58

prospective studies. Lancet 2011 377 1085-1095.

29. Meigs JB, Wilson PW, Fox CS, Vasan RS, Nathan DM, Sullivan LM & D’Agostino

RB. Body mass index, metabolic syndrome, and risk of type 2 diabetes or

cardiovascular disease J Clin Endocrinol Metab 2006 91 2906–2912.

30. Soriguer F, Gutiérrez-Repiso C, Rubio-Martín E, García-Fuentes E, Almaraz MC,

Colomo N, Esteva de Antonio I, de Adana MS, Chaves FJ, Morcillo S, Valdés S &

Rojo-Martínez G. Metabolically healthy but obese, a matter of time? Findings from

the prospective Pizarra study. J Clin Endocrinol Metab 2013 98 2318-2325.

31. Godsland IF. Intra-individual variation: significant changes in parameters of lipid and

carbohydrate metabolism in the individual and intra-individual variation in different

test populations. Ann Clin Biochem 1985 22 618-624.

32. Katzmarzyk PT, Janssen I, Ross R, Church TS & Blair SN. The importance of waist

circumference in the definition of metabolic syndrome: prospective analyses of

mortality in men. Diab Care 2006 29 404–409.

33. Song Y, Manson JE, Meigs JB, Ridker PM, Buring JE & Liu S. Comparison of

usefulness of body mass index versus metabolic risk factors in predicting 10-year risk

of cardiovascular events in women. Am J Cardiol 2007 100 1654–1658.

34. St-Pierre AC, Cantin B, Mauriège P, Bergeron J, Dagenais GR, Després JP &

Lamarche B. Insulin resistance syndrome, body mass index and the risk of ischemic

heart disease. . CMAJ 2005 172 1301–1305.

35. Calori G, Lattuada G, Piemonti L, Garancini MP, Ragogna F, Villa M, Mannino S,

Crosignani P, Bosi E, Luzi L, Ruotolo G & Perseghin G. Prevalence, metabolic

features, and prognosis of metabolically healthy obese Italian individuals: the

Cremona Study. Diab Care 2011 34 210–215.

36. Ogorodnikova AD, Kim M, McGinn AP, Muntner P, Khan U & Wildman RP. Incident

cardiovascular disease events in metabolically benign obese individuals. Obesity

(Silver Spring) 2012 20 651–659.

37. Flint AJ, Hu FB, Glynn RJ, Caspard H, Manson JE, Willett WC & Rimm EB. Excess

weight and the risk of incident coronary heart disease among men and women.

Obesity 2010 18 377–383.

Table 1: Cohort characteristics

nobservations

Median (IQR) /% (n)

Age (y) 1191 46.4 (39.9-51.8)BMI (kg/m2) 1182 25.6 (23.7-27.4)

BMI ≥27.0 kg/m2 % (n) 1182 30 (350)

Fasting plasma glucose (mmol/L) 1177 5.3 (5.0-5.6)

Fasting plasma insulin (mmol/L)* 1030 9.5 (6.0-14.5)

SBP (mm Hg) 1190 123 (115-135)

DBP (mm Hg) 1190 80 (70-85)

Total Cholesterol 1181 5.36 (4.71-6.11)

Triglycerides (mmol/L) 1181 1.22 (0.86-1.74)

HDL cholesterol (mmol/L) 1004 1.25 (1.08-1.43)

LDL cholesterol (mmol/L) 1003 3.50 (2.86-4.14)

Uric acid (µmol/L) 1163 380 (326-430)

HOMA-IR 1030 2.19 (1.36-3.41)

Matsuda-IS 699 4.71 (3.07-7.17)

White cell count (x109) 699 4.71 (3.07-7.17)

ESR 949 5.6 (4.7-6.8)

Family History of Diabetes % (n) 1181 35 (411)

Family History of Heart Disease % (n) 1181 8 (93)

Existing diabetes % (n) 1171 5 (57)

Smoking % (n) 1184

Non-smoker 67 (797)

<15 Cigarettes/d 20 (241)

≥15 Cigarettes/d 12 (146)

Alcohol % (n) 1179

<28 U/week 64 (753)

≥ 28 U/week 36 (426)

Exercise % (n) 1126

No exercise 41 (460)

Nonaerobic 41 (457)

Aerobic 18 (209)

Current Drug Use % (n) Glucose-lowering 1191 0.3 (4)

Lipid-lowering 1191

Statin 0.3 (4)

Fibrate 0.3 (3)

Blood pressure-lowering 1191 2.2 (26)

Uric acid-lowering 1191 1.0 (12)

Mortality Total mortality % (n) 1164 29 (340)

CVD mortality % (n) 1164 10 (111)

mortality follow-up time (years) 1176 24.8 (18.6-32.9)

Table 2: Characteristics of individuals metabolically healthy (MH) or metabolically unhealthy (MUH) according to conventional criteria among normal

weight (NW) and overweight or obese (OWOB) participants. Median (IQR) for continuous variables, percentage (n) for categorical variables. Significant

differences between NW and OWOB in MH and MUH groups are shown.

MH p MUH pNW (n=425) OWOB (n=90) NW (n=358) OWOB (n=226)

age (years) 44.4 (37.6-49.7) 45.4 (39.7-52.0) 0.09 46.9 (42.3-53.4) 48.0 (42.5-53.8) 0.7

BMI (kg/m2) 24.2 (22.5-25.4) 28.2 (27.8-29.3) <0.001 24.8 (23.8-26.0) 28.7 (27.7-30.6) <0.001

Fasting plasma glucose (mmol/L) 5.1 (4.9-5.3) 5.3 (5.0-5.5) 0.001 5.4 (5.1-5.8) 5.5 (5.2-5.9) 0.1

Fasting plasma insulin (mmol/L) 7.5 (5.0-11.5) 11.0 (7.5-15.5) <0.001 9.2 (6.0-14.5) 13.0 (9.5-18.5) <0.001

SBP (mm Hg) 120 (110-124) 120 (110-130) 0.004 130 (120-140) 130 (120-140) 0.3

DBP (mm Hg) 70 (70-80) 80 (70-84) <0.001 80 (75-90) 85 (80-90) 0.004

Total Cholesterol 4.9 (4.4-5.4) 4.8 (4.4-5.4) 0.5 5.9 (5.3-6.4) 5.8 (5.2-6.6) 0.7

Triglycerides (mmol/L) 0.96 (0.72-1.23) 1.11 (0.76-1.46) 0.007 1.45 (1.04-1.98) 1.78 (1.31-2.55) <0.001

HDL cholesterol (mmol/L) 1.38 (1.20-1.55) 1.26 (1.14-1.43) 0.002 1.22 (1.04-1.40) 1.12 (0.96-1.27) <0.001

LDL cholesterol (mmol/L) 3.01 (2.55-3.59) 2.95 (2.60-3.32) 0.6 3.89 (3.43-4.40) 3.72 (3.27-4.32) 0.1

Uric acid (µmol/L) 350 (304-400) 386 (340-449) <0.001 380 (330-420) 410 (359-460) <0.001

HOMA-IR 1.67 (1.08-2.67) 2.63 (1.79-3.65) <0.001 2.27 (1.41-3.47) 3.29 (2.26-4.68) <0.001

Matsuda-IS 6.03 (4.29-8.24) 3.87 (2.84-4.78) <0.001 4.37 (3.04-6.66) 2.70 (1.90-4.23) <0.001

White cell count (x109) 5.2 (4.5-6.3) 6.1 (5.2-7.0) <0.001 5.7 (4.8-7.1) 5.8 (5.0-6.9) 0.2

ESR 3 (2-6) 4 (2-6) 0.08 5 (2-8) 5 (2-10) 0.4

FH heart disease, 1° relative % (n) 33 (141) 26 (23) 0.3 38 (135) 37 (83) 0.2

FH diabetes, 1° relative % (n) 8 (32) 8 (7) 0.9 8 (29) 8 (17) 0.8

diabetes % (n) 1 (5) 1 (1) 0.9 5 (19) 5 (12) 0.9

Smoking % (n) 0.009 0.3

Non-smoker 74 (313) 60 (53) 65 (231) 71 (159)

<15 Cigarettes/d 16 (67) 29 (26) 22 (78) 18(40)

≥15 Cigarettes/d 10 (44) 11 (10) 13 (46) 12 (26)

Alcohol % (n) 0.002 0.8

<28 U/week 75 (314) 59 (53) 59 (209) 58 (132)

≥ 28 U/week 25 (106) 41 (37) 41 (145) 42 (94)

Exercise % (n) 0.2 0.1

No exercise 34 (138) 36 (31) 41 (138) 50 (111)

Nonaerobic 43 (173) 49 (43) 41 (137) 36 (79)

Aerobic 23 (95) 15 (13) 18 (59) 14 (32)

Current Drug Use % (n) Glucose-lowering 0 (0) 1 (1) 0.03 0.9 (3) 0.4 (1) 0.5

Statin 0 (0) 0 (0) 1.0 0.3 (1) 0.9 (2) 0.3

Fibrate 0 (0) 0 (0) 1.0 0 (0) 0.4 (1) 0.2

Blood pressure-lowering 0.5 (2) 1.1 (1) 0.4 2.2 (8) 4.0 (9) 0.2

Mortality Total mortality % (n) 18 (77) 23 (21) 0.2 34 (122) 31 (71) 0.5

CVD mortality % (n) 5 (21) 2 (2) 0.2 10 (35) 13 (29) 0.2

mortality follow-up time (years) 24.9 (18.7-32.9) 22.3 (17.4-29.5) 0.008 25.4 (18.5-32.4) 22.5 (17.4-28.7) 0.002

Table 3: Proportions of individuals normal weight (NW) or overweight or obese (OWOB) and metabolically healthy at baseline remaining metabolically healthy at successive visits. Number, n, are numbers at first visit who were MH

1+ visits 2+ visits 3+ visits 4+ visits 5+ visits 6+ visits 7+ visits 8+ visits 9+ visits 10+ visits n-weighted averagen 425 90 225 41 153 30 108 25 77 17 59 16 46 15 32 11 22 7 9 4

NW OWOB NW OW

OB NW OWOB NW OW

OB NW OWOB NW OW

OB NW OWOB NW OW

OB NW OWOB NW OW

OB NW OWOB

percentage remaining metabolically healthyVisits

1 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 1002 71§ 49§ 68§ 57§ 67§ 60§ 68§ 59† 70§ 57† 66§ 54† 60§ 55* 55§ 43* 67 25* 68 543 53* 44 50* 44 50* 48 55 44 53 47 44 46 41 29 45 25 51 484 40 36 38 36 45 32 42 34 41 46 41 29 45 25 41 395 28 30 31 25 29 27 32 37 41 15 45 0 31 236 29 25 29 27 32 37 41 15 45 0 32 337 22 27 22 37 28 15 34 0 24 198 16 28 19 15 23 0 18 149 10 0 12 0 11 0

10 12 0 12 0

Proportions compared with preceding visit - significances: § p<0.001; † p<0.01; * p<0.05

Table 4. Weighted average percentages of individuals metabolically healthy at baseline remaining metabolically healthy. Data for up to 3 visits subsequent to the baseline visit are shown, with baseline visits taken as an individual’s first, second or third visit. Weighted average percentages derive from subsets of individuals with 2, 3, 4 to 10 or more visits.

% remaining metabolically healthy2nd visit 3rd visit 4th visit

Baseline visit Percent remaining metabolically healthyNormal weightvisit 1 68 51 41visit 2 68 47 38visit 3 66 54 49Overweight or obesevisit 1 54 48 39visit 2 70 55 39visit 3 56 39 41

Legend to Figures

Figure 1.

Proportions of individuals overweight or obese (OWOB - dashed line) or normal weight (NW

- solid line) and metabolically healthy at baseline remaining metabolically healthy. Attrition of

metabolic health is illustrated by plotted series of observations for subgroups that had 6 or

more, 7 or more or 8 or more visits.

Supplementary Table 1: Criteria for metabolic health as predictors of cardiovascular disease (CVD) and total mortality in unadjusted Cox proportional hazards models after exclusions for missing values and existing CVD (n=1099).

CVD mortality Total mortality

Metabolic risk factors hazard ratio(95% CI) p hazard ratio

(95% CI) p

plasma triglyceride ≥1.7 mmol/lor use of fibrates 1.82 (1.16-2.86) 0.009 1.44 (1.12-1.87) 0.005

HDL cholesterol <1.03 mmol/l 1.79 (1.07-3.02) 0.02 1.50 (1.11-2.02) 0.008

blood pressure ≥130 mmHg systolicor ≥85 mmHg diastolicor use of blood pressure-lowering agents

2.69 (1.68-4.31) <0.001 1.86 (1.46-2.36) <0.001

fasting plasma glucose ≥5.6 ordiagnosed diabetesor use of glucose-lowering agents 1.69 (1.08-2.66) 0.02 1.64 (1.28-2.10) <0.001

LDL-C ≥3.36 mmol/lor use of statins 1.92 (1.21-3.04) 0.005 1.33 (1.05-1.68) 0.01

Number of metabolic risk factors1 1.66 (0.62-4.48) 0.3 1.51 (0.95-2.40) 0.082 2.79 (1.07-7.27) 0.03 2.07 (1.31-3.28) 0.0023 4.47 (1.72-11.6) 0.001 3.02 (1.91-4.80) <0.0014 8.12 (2.57-25.7) <0.001 4.20 (2.24-7.89) <0.0015 43.4 (11.5-164.0) <0.001 11.0 (4.14-28.9) <0.001

Supplementary Table 2: Percent prevalence of MH at successive visits in individuals who were NW or OWOB at first visit and who had 1 or more through to 10 or more visits. Number, n, are numbers at first visit who were MH

1+ visits 2+ visits 3+ visits 4+ visits 5+ visits 6+ visits 7+ visits 8+ visits 9+ visits 10+ visitsn 783 316 448 160 308 96 213 71 159 51 122 46 96 40 69 32 46 17 23 9

NW OWOB NW OWOB NW OWOB NW OWOB NW OWOB NW OWOB NW OWOB NW OWOB NW OWOB NW OWOB

Visits 1 54 28 50 26 50 31 51 35 48 33 48 35 48 38 46 34 48 41 39 44

2 50 29 47 31 48 38 49 42 52 41 51 39 48 41 40 46 44 433 49 34 46 36 43 46 45 47 47 48 45 53 38 54 36 714 44 31 42 34 44 31 47 35 46 50 50 56 55 675 39 37 41 36 40 33 44 45 49 50 52 576 43 33 43 31 45 38 49 50 61 337 38 28 41 33 45 43 54 508 34 33 36 50 45 309 30 37 35 33

10 33 38